Compared to online, or real-time video coding, offline, or non-real-time video coding has two major advantages on coding: (i) the whole video clip or sequence is available in advance, and (ii) no stringent coding time constraint. The 1st advantage implies joint, or global, optimization of coding of all the frames altogether, i.e. globally optimized FBA, and the 2nd advantage implies that encoding of a clip can be conducted multiple times, or passes. Both will lead to greatly improved overall video coding performance.
At least one implementation described herein addresses an important problem in sequence-wise globally optimized FBA for offline video coding, i.e. frame-level QP offset calculation. In most recent rate control solutions, accurate R-QP modeling is a commonly adopted approach for carrying out effective GOP-, frame-, and MB-level bit allocations. In the case of FBA, the general optimization objective is to achieve constant coding quality for all the concerned frames while satisfying the total bit budget constraint and specific buffer constraint. Usually, to fulfill the constant quality request, a same value of QP will be assumed for all the frames when calculating their respective allocated bits. Moreover, a particular QP offset has to be calculated for each frame respectively to guarantee a good overall coding performance at last.